Introduction and methods
Assessing the environmental burden of disease at national and local levels
Annette Prüss-Üstün Colin Mathers Carlos Corvalán Alistair Woodward
Series Editors
Annette Prüss-Üstün, Diarmid Campbell-Lendrum, Carlos Corvalán, Alistair Woodward
World Health Organization
Protection of the Human Environment Geneva 2003
introduction and methods / edited by Annette Prüss-Üstun … [et al.]
(Environmental burden of disease series ; no.1)
1.Environmental health 2.Environmental exposure 3.Risk factors 4.Cost of illness 5.Disability evaluation 6.Health status indicators 7.Risk assessment - methods 8.Manuals I.Prüss-Üstün, Annette. II.Series.
ISBN 92 4 154620 4 (NLM Classification: WA 30)
ISSN 1728-1652
Suggested citation
Prüss-Üstün A, et al. Introduction and methods: assessing the environmental burden of disease at national and local levels. Geneva, World Health Organization, 2003. (WHO Environmental Burden of Disease Series, No. 1).
© World Health Organization 2003
All rights reserved. Publications of the World Health Organization can be obtained from Marketing and Dissemination, World Health Organization, 20 Avenue Appia, 1211 Geneva 27, Switzerland (tel:
+41 22 791 2476; fax: +41 22 791 4857; email: bookorders@who.int). Requests for permission to reproduce or translate WHO publications – whether for sale or for noncommercial distribution – should be addressed to Publications, at the above address (fax: +41 22 791 4806; email:
permissions@who.int).
The designations employed and the presentation of the material in this publication do not imply the expression of any opinion whatsoever on the part of the World Health Organization concerning the legal status of any country, territory, city or area or of its authorities, or concerning the delimitation of its frontiers or boundaries. Dotted lines on maps represent approximate border lines for which there may not yet be full agreement.
The mention of specific companies or of certain manufacturers’ products does not imply that they are endorsed or recommended by the World Health Organization in preference to others of a similar nature that are not mentioned. Errors and omissions excepted, the names of proprietary products are distinguished by initial capital letters.
The World Health Organization does not warrant that the information contained in this publication is complete and correct and shall not be liable for any damages incurred as a result of its use.
Table of Contents
Preface ... v
Affiliations and acknowledgements ... vi
Summary... vii
1. The WHO guides on assessing the environmental burden of disease ... 1
1.1 Objective of the guides ... 1
1.2 Target readership ... 2
1.3 Content of this series ... 2
1.4 Adapting the guides to specific needs ... 2
1.5 Improving the evidence base ... 2
1.6 What evidence is missing? ... 3
1.7 Related activities... 3
2. Background to assessing the environmental burden of disease... 4
2.1 Why measure the EBD? ... 5
2.2 What are environmental risk factors and how are they categorized? ... 9
2.3 Attributable risk... 12
2.4 Limitations of EBD studies ... 16
2.5 What are the links between EBD assessments and policy-making? ... 17
2.6 Data and indicators for EBD assessments ... 20
3. The Global Burden of Disease concept ... 27
3.1 Introduction ... 27
3.2 Summary measures of population health... 27
3.3 Quantifying time lived with disability... 28
3.4 Other social values... 30
3.5 Calculation of DALYs with discounting and age weighting... 32
3.6 Relating summary measures of health to the causes of loss of health... 35
3.7 The GBD 2000 study – an analysis of global mortality patterns... 36
3.8 The GBD 2000 study – epidemiological analyses for calculating YLD ... 36
3.9 Main findings from the GBD 2000 study... 36
4. Methods for estimating the environmental burden of disease... 41
4.1 General method... 41
4.2 Alternative or counterfactual exposure... 45
4.3 Choosing the study population for an EBD assessment ... 47
4.4 Estimating the EBD when NBD data are available... 48
4.5 Estimating the EBD from disease-specific national health statistics when NBD data are not available... 48
4.6 Estimating the EBD from limited national or local health statistics when NBD data are not available... 49
4.7 Estimating the EBD using preliminary NBD estimates from WHO... 50
4.8 Estimating the EBD for diseases that are caused by one risk factor ... 51
4.9 Estimating the EBD for diseases not assessed by national statistics or by WHO... 52
4.10 Estimating uncertainty... 53
References ... 57
Glossary of terms for the EBD series ... 60
Preface
To prevent disease and injury it is essential that their underlying causes (health risks) are quantitatively attributed. Together with information on the costs of interventions, their effectiveness and the socioeconomic context, such knowledge provides a rational basis for policy-setting. While quantitative studies have been performed for some health risks, few have assessed the disease burden from environmental risk factors and, traditionally, the studies have focused on a single risk factor. Recently, however, WHO analysed the global burden of disease from 26 risk factors, and the results were published in the World Health Report 2002 (WHO, 2002). The Environmental Burden of Disease (EBD) series of guides is based on the same methodological framework as used in the World Health report, and provides practical guidance on assessing the health impacts of environmental risk factors.
The guides, together with accompanying material, such as the spreadsheets available for certain risk factors at web site www.who.int/phe, should provide sufficient methodological information to perform the EBD assessments.
The EBD series of guides is composed of an introductory volume, and volumes that provide detailed guidance for assessing the health burden of specific environmental risk factors. Most of the guides focus on assessments of national and local populations, which are most relevant for policy-making. In some volumes, however, the global disease burden is assessed for certain health risks. All the guides take a practical, step-by-step approach and use numerical examples. The methods described in the guides can be adapted both to local and national levels, and can be tailored to suit data availability. In this introductory volume, the methodological framework for quantitatively assessing health impacts at population level is described. It is recommended that the framework be adopted by other EBD studies, to ensure that estimates are both reliable and comparable.
Affiliations and acknowledgements
Annette Prüss-Üstün, Colin Mathers, Carlos Corvalán and Diarmid Campbell-Lendrum are from the World Health Organization, and Alistair Woodward is from the Wellington School of Medicine, New Zealand.
We would like to acknowledge the many experts around the world who, over several years, have contributed to the development of methods for estimating the disease burden of environmental risk factors. In particular, we would like to thank participants of the meeting, Methodology for assessment of environmental burden of disease (held in Buffalo, NY, USA in 2000), and participants of two regional meetings on the environmental burden of disease (Curitiba, Brazil in 2002; and Damascus, Syria in 2002). Peer reviewers of the sections have also provided invaluable comments.
The financial support of the US Environmental Protection Agency is also greatly appreciated. We gratefully acknowledge the editing by Kevin Farrell, and layout by Eileen Brown, who have put this document into its final form.
Abbreviations used BoD Burden of disease.
EBD Environmental burden of disease.
GBD Global burden of disease.
NBD National burden of disease.
YLL Years of life lost due to premature mortality.
YLD Years lived with disability.
Summary
This introductory guide provides the background to, and a description of, the general method for assessing the disease burden caused by environmental risk factors. Subsequent guides address the disease burdens of specific environmental risk factors. To assess a disease burden, the health impact of disease and injury needs to be assessed quantitatively at population level. This may be measured in terms of the number of deaths, or as a summary measure of population health, such as the disability-adjusted life year (DALY).
Environmental burden of disease (EBD) studies assess the disease burden attributable to environmental risk factors, and are closely linked to assessments of the disease burden for individual diseases and injuries. Indeed, the burden of disease from disease and injury has been assessed at global level, and national level data are becoming available, which can be used in EBD studies. The results of disease burden studies are generally presented by gender and by age group, and are measured in terms of deaths and DALYs. The actual calculations for an EBD assessment are relatively simple once the input data (exposure and health outcomes) have been collected in a suitable format. The method can also be adapted to the health statistics that are available for the study population.
EBD assessments do not necessarily entail large costs. In many countries and regions, environmental health indicators are already routinely assessed, but are not yet processed into health information. Certain of these indicators can be used directly as input for EBD assessments, so additional assessments may not be necessary. The accuracy of EBD assessment will, however, depend on the quality of the data used as input.
Attributing the health impacts of environmental risk factors at population level can serve several public health activities. It can help to prioritize actions for preventing or reducing health impacts in the population, and by allowing the future health burden to be estimated, an EBD assessment can inform planning for preventive action. EBD assessments can also be used to estimate performance indicators for health-supporting environments, and identify high-risk groups in the population. Finally, EBD information can also be used to predict the health gains that interventions (including regulations) will bring to a population.
1. The WHO guides on assessing the environmental burden of disease
This guide is the first in a series about estimating the disease burden of environmental risk factors. It provides an introduction to the environmental factors that pose a risk to health, and outlines the general methods used to estimate the disease burden of these factors. It also introduces the Global Burden of Disease (GBD) concept (Murray & Lopez, 1996), describes National Burden of Disease (NBD) studies (Mathers et al., 2001) and provides a summary of environmental health indicators.
Other guides in the series focus on specific risk factors and on how to assess the associated disease burdens (Box 1). It is hoped that the guides will help to strengthen local capacity in the analysis and interpretation of environmental health data, and assist with decision- making at national level.
a The disease burden of the risk factors listed to the left in Box 1 has been assessed at global level (WHO, 2002), together with that of 16 other risk factors from areas such as lifestyle, diet- related risks, use of addictive substances, unsafe sex and unsafe health practices (Ezzati et al., 2003).
1.1 Objective of the guides
The objective of the guides is to provide practical information to countries on how to assess what fraction of a national or subnational disease burden is attributable to an environmental risk factor. To assess the disease burden of a risk factor, the harmful effects of the risk factor on human health must be estimated fully, as well as the distribution of the harmful effects in the population. Any estimates and assumptions used in the assessment should be stated explicitly. The outcome of the assessment is information that can be used: to guide policies and strategies both in the health sector and in the environmental sector; to monitor health risks; and to analyse the cost-effectiveness of interventions. For example, the information can highlight the contribution of major environmental risk factors to the total disease burden of a country or study population. Or, it can be used to estimate changes in the disease burden and avoidable disease burden, following interventions to reduce an environmental risk factor or to change behaviour.
· Ambient air
· Indoor air
· Lead
· Water, sanitation and hygiene
· Climate change
· Occupational factors:
- injuries - noise - carcinogens - dusts
- ergonomic stressors - sharps injuries in health-care
workers
· Nutrition
· UV radiation
· Recreational water-quality
· Fluoride in drinking-water
· Arsenic in drinking-water
· Nitrates in drinking-water
· Community noise
· Poverty Box 1: Risk factors covered in the guidesa
More generally, an assessment of the environmental burden of disease (EBD) can be used to raise awareness and strengthen institutional capacity for reducing the impact of environmental health risks on the population. The EBD can be assessed for an entire country, or applied to the subnational level (e.g. a city or district), provided basic data are available for the chosen perimeter. EBD studies complement NBD studies, as well as studies of other behavioural or physical risk factors, such as alcohol intake, high levels of blood cholesterol, and unsafe sex.
1.2 Target readership
The target readership includes professionals, such as researchers in universities, government agencies or the private sector, and decision-makers at national or regional level, who are interested in quantitatively estimating the health impacts of environmental risk factors.
1.3 Content of this series
The Introduction (Sections 1–4) addresses the relevance of an EBD assessment to policy, as well as to the larger framework of environmental health assessment, management and evaluation. The general method for assessing the disease burden of a risk factor is also critically reviewed, as are alternative methods, and specific issues are evaluated, such as the units of measurement of disease burden, health valuation, and discounting of future outcomes. There is some guidance on adapting the methods to local needs and circumstances.
Further volumes in the series are composed of guides that help professionals quantify the EBD from some specific environmental risk factors. In these volumes, practical steps are described for performing the quantitative assessments of risk, and for processing the risk data into burden of disease (BoD) data. Data requirements are also given.
Of particular concern are the uncertainties around estimates and the interpretation of results. These issues are addressed in Section 4, and also in the volumes on specific risk factors. In the Annexes, global EBD estimates are given for 10 major environmental risk factors, including 5 occupational risk factors. The data are shown by region, by gender and for 8 age groups.
1.4 Adapting the guides to specific needs
The guidance provided in the series can be adapted to a country’s specific needs, to available data sets, or to the desired degree of accuracy. If more locally-specific data, or new exposure-response relationships become available, these can also be used to complement or update the evidence given in the guides for the various risk factors.
1.5 Improving the evidence base
progress that has been made is that, two decades ago, it probably would not have been possible to implement the EBD methods proposed in this series.
A developing evidence base also means that the methods proposed in the EBD guides should be updated as new links between health and the environment are uncovered. The new information could help to improve the accuracy of quantitative linkages between health and the environment, or improve the geographical applicability of data, or better describe the health impacts on poorly-assessed subgroups in a population (e.g. women, or people in a particular age range).
1.6 What evidence is missing?
Although current evidence on the relationship between exposure and disease is solid enough to develop quantitative estimates of the disease burden for a number of environmental risk factors, many other risk factors have not been well documented. In particular, it is easy to overlook risk factors with long latency periods or nonspecific outcomes; factors with exposures that are difficult to assess at population level; and factors that are distal to the outcome. And the absence of data does not necessarily mean that the BoD is negligible or absent. The results of risk factor assessments should therefore be interpreted with caution, and BoD assessments should be regarded as the best current estimates of the magnitude of health problems due to environmental factors.
1.7 Related activities
An assessment of the BoD by risk factor (which includes an EBD assessment) is closely linked to an assessment of the BoD by disease. Indeed, an EBD assessment is best performed after a NBD study has been developed (Mathers et al, 2001), but a prior NBD assessment is not essential for an EBD assessment.
2. Background to assessing the environmental burden of disease
An assessment of the BoD quantifies the amount of disease at population level. Ideally, the assessment should be carried out in an internally consistent way, and use common units of measurement, since this will allow data on the disease burden and risk factors to be compared between studies. Assessments carried out in this way would also allow the data to be compared for different population groups and across geographical regions. For this reason, a summary measure of population health is used in assessments, which serves as a common currency. Although various summary measures have been developed, the disability-adjusted life year (DALY) is most frequently used in this text. This measure combines the number of years of healthy life lost due to premature mortality and to disability. A detailed explanation of how the DALY is calculated is given in Section 3).
Traditional assessments of the number of healthy years lost have measured either the number of deaths due to disease, or the disease incidence, but not both, which makes it difficult to compare losses that occur at different ages, or from different causes of ill-health.
How does a death at age 20 years compare with a death at age 70 years? How do 200 acute respiratory infections compare to 400 cases of infectious diarrhoea? Summary measures of population health, such as the DALY, provide a framework for dealing with these difficult questions.
Assessments are internally consistent, provided that similar approaches are taken across diseases, risk factors and geographical regions. And the assessments should be coherent – for example, the total number of deaths estimated by a BoD study should not exceed the total number of deaths registered in a country. Death registers generally provide the most complete set of health statistics, but specific causes of death may be recorded in different ways. Many countries have carried out national BoD studies, which estimate the amount of ill-health (commonly measured in DALYs) that is attributable to different disease categories. These studies provide a basis for estimating the EBD, and provide an opportunity to compare health losses due to different risk factors or disease states. In such cases, the internal consistency is already ensured to a large extent. However, a national BoD study is not essential, and the fraction of deaths or morbidity caused by environmental factors can still be estimated without such data.
Main issues
· A BoD study quantifies the health gap at population level and can form the basis of an EBD assessment;
· Summary measures of population health make it possible to compare different estimates of the EBD by standardizing methodology;
· Estimates of the EBD should be internally consistent and
2.1 Why measure the EBD?
This section explores the rationale for assessing disease burden and outlines ways in which such information can contribute to policy formulation. An important reason for using the formal EBD approach is that it is open to scrutiny. The scientific input to policy decisions often includes important assumptions and judgments that are unspoken (or unwritten). This lack of clarity is why knowledgeable people, when faced with common problems, frequently come to different conclusions. The EBD approach helps to understand the reasons for divergent opinion, without which it will be difficult to develop effective policies.
For various reasons, the magnitude of the health problem from environmental factors needs to be estimated. Health ministries, researchers, bodies responsible for setting standards, scientific advisory groups, and international aid organizations all require such estimates.
The questions posed in assessments of the health burden may seem relatively straightforward (e.g. “what is the impact of unvented house fires on the health of families?”), but the answers are often difficult to pin down. For example, it may not be immediately obvious how the exposures and outcomes should be defined (what is meant by the “health of families”?), or what time scale to use (are we dealing with current effects of past exposures, or future effects of current exposures?). The alternative scenario may also be unclear (what is the impact of unvented fires to be compared to? Open fires with chimneys? Wood-burning stoves? No fires at all?). Difficulties also arise in defining health outcomes (should asthma, pneumonia and ear infections be lumped together, or counted separately? Should deaths and nonfatal illnesses be combined to give a single measure of “impact”, and how?).
The EBD method provides a formalized, explicit approach, in which the choices of inputs are apparent. This allows the effects of different assumptions to be readily displayed (e.g.
DALYs due to indoor air pollution from unvented fires can be calculated with and without age-weighting). In this way, the EBD approach is not just a tool for improving our understanding of environmental and health linkages – it also allows estimates from different sources to be compared and communicated in a standardized format. The steps for calculating the EBD are described in more detail in Section 4.
There are several other good reasons for performing EBD studies, including:
Prioritizing actions in health and the environment
EBD information supports decisions on priority actions in health and the environment. A common problem in both developed and developing countries is that resources are limited, and informed choices about health have to be made under circumstances where it may not be possible to achieve “safe environmental levels” of every known hazardous substance, given the available resources. For example, air pollution levels are regularly exceeded in most major urban centres, but available resources are not always adequate for modifying transport policies or technologies. Many beaches also do not meet the standards for bathing-water, but improved sewage treatment may not be affordable in the short term.
EBD assessments do not replace decision-making in environmental health, but are designed to assist in the process of weighting the advantages and disadvantages of alternative interventions. The purpose is not to provide a fine-tuned calculus of priorities, but instead, an indication of the relative effects of environmental exposures.
Planning for preventive action
When planning to prevent or reduce problems associated with a high disease burden, the information provided by an EBD study is essential for prioritizing actions. EBD data about the effectiveness and cost-effectiveness of interventions (i.e. the dollar cost per unit reduction in environmental exposures or DALYs) will indicate the preventability and the relative costs of the disease burden, respectively. EBD data can also inform preventive actions by being used as input for infrastructure planning, for example, or for estimates of the future disease burden.
However, information other than EBD data also needs to be taken into account when planning preventive interventions. The social and ethical context, for example, should be considered systematically. Social considerations could include the ability of people to respond to their own needs, the social consequences of disease burden, and the priority placed on reducing health inequalities.
Assessing performance
Data from an EBD study can be used to calculate performance indicators for health- supporting systems and environments in a country or region. If a study has the necessary resolution, it can map out geographical or population-specific differences, and monitor trends. The performance indicators can be used to compare the developmental status of regions; they can be compared with other measures of the developmental status of a region;
or can be used to compare regions of similar developmental status.
Comparing action and health gain
EBD information provides the opportunity to manage environmental risks from a new perspective. For example, since available resources do not always allow the risks from environmental exposure to be reduced to zero, government agencies might concentrate on risks that provide the best opportunities for gain, for a given investment of resources. EBD information is useful in such an evaluation, and allows the actual health gain of an action in environmental management (or related behaviour change) to be estimated. While most environmental health guidelines attempt to answer the question, “at which value can we reasonably expect that no observable health impacts will occur in an exposed population,”
the EBD assessments in the guides answer the following:
- which of the environmental burdens generate the largest impacts on public health;
- by how much will the disease burden in a population be reduced if guidelines are implemented;
- which reductions in exposures would generate the greatest change in DALYs for a given cost;
- what is the cheapest way to achieve a given reduction in DALYs.
Identifying high-risk populations
An EBD study, or more generally, a BoD study, can identify the important contributions to health inequalities in high-risk populations. Although routine health statistics (e.g.
mortality registrations) may point to population subgroups at high risk (e.g. women, the elderly, people low on the socioeconomic scale or ethnic groups), EBD estimates can help to understand the causes of inequalities in the subgroups by attributing health gaps to particular environmental exposures, and help to direct public health efforts accordingly (Box 2.1).
Planning for future needs
It is possible to project exposures into the future and estimate trends in the EBD, even when there may be a long time-lag between exposure and the onset of disease. Such estimates of
Box 2.1 Adjusting burden of disease estimates for equitya
In 1999, a NBDb study was carried out in New Zealand to identify the impact of over 100 major diseases and injuries, and 8 chronic disease risk factors. Two years later, a further analysis was conducted, to examine more closely the distribution of the disease burden between indigenous (Maori) and nonindigenous ethnic groups. In general, the health of non-Maoris is considerably better than that of Maoris. Age- specific mortality of non-Maoris, for instance, is approximately half that for Maoris.
The conventional DALY approach was modified using an “impact share” model. This takes into account the variation in the distribution of the BoD between Maori and non- Maori, and the extent to which that contributes to the total Maori/non-Maori gap in health status.
For each disease or risk factor, an equity adjustment factor was calculated as:
1+ [p(RR-1)/(p(RR-1)+1)]
where RR is the age-standardized DALY rate ratio (diseases) or prevalence rate ratio (risk factors), and p is the proportion of the total difference between Maori and non- Maori age-standardized DALY rates accounted for by the condition of interest. (See the Glossary for a definitions of these terms.)
The equity factor was multiplied by the number of DALYs lost for each disease or risk factor. The purpose of the adjustment was to assist policy-makers in setting priorities for health expenditure. The effect of including the adjustment for equity was to boost the relative importance of diseases and risk factors that weigh disproportionately on the health of Maori. This was apparent when the top 25 causes of DALYs lost in New Zealand were compared, with and without the adjustment for equity. Differences included:
- diabetes was ranked higher;
- a lack of physical activity was ranked lower as a risk factor;
- sudden infant death syndrome was included on the list;
- hearing disorders were no longer in the top 25 causes of disease.
a Source: New Zealand Ministry of Health (2001)
the future burden give policy makers the option of shifting priorities proactively, if this is likely to be less costly than doing nothing until the disease burden is apparent. For example, on the basis of past exposures it is possible to project future trends in mesothelioma and other asbestos-related conditions. Moreover, the future EBD can be estimated for asbestos exposures, depending on different control strategies.
Assessing future scenarios
There have been several recent assessments of future global changes. For example, the Intergovernmental Panel of Climate Change produced a set of CO2 emission scenarios that could lead to global temperature increases, and the United Nations Environment Programme produced a set of environmental scenarios in its Global Environmental Outlook. Such scenarios can be used to paint a picture of future population exposures, and to estimate health losses that may occur. Assessments of this kind carry a high level of uncertainty, but they do provide indications of what may happen if certain environmental conditions apply. Forward-looking assessments like these are important with problems such as climate change, where there are long time-lags between emission of greenhouse gases, changes in the climate, and subsequent health effects.
Setting priorities in health research
To some extent, an EBD assessment at national level can also foster priority-setting in health research. At present, resources spent on research and development do not always reflect the disease burden. This is apparent from the fact that only 10% of the total resources spent on health research is devoted to the health problems of 90% of the world’s population (WHO, 1996). The following five criteria have been proposed for making rational decisions about allocating resources for research (WHO 1996):
1. what is the attributable BoD (this is addressed in this guide);
2. what are the main determinants of the BoD and its persistence;
3. what is the knowledge base of the disease (including the cost-effectiveness of interventions to reduce the disease burden);
4. what is the likelihood that cost-effective interventions can be developed;
5. what are the present resource flows for the risk factor or disease.
Relevancy to policy-making
EBD assessments may be more relevant than disease-based assessments for decision- making in environmental health policies, since the health gain from changes in environmental exposures can be directly estimated with these methods. The links between EBD information and policy are discussed in Section 2.5.
2.2 What are environmental risk factors and how are they categorized?
Environmental causes of disease may be categorized in many ways, e.g. by referring to media which may carry hazards, as individual risk factors (agents), or according to the nature of the hazard.
Media that carry hazards include:
- water used for drinking, recreational activities or agricultural activities (such as irrigation);
- food;
- special environments that potentially carry hazards, such as agricultural environments, water resources, or wetlands;
- indoor and outdoor air.
Individual risk factors include:
- chemical substances;
- noise;
- radiation (ionizing, UV, electromagnetic).
These risk factors can be further divided into those in the occupational environment, or in the general environment (i.e. non-occupational environment). Many of the media risk factors and individual risk factors overlap. Risk factors also present different types of hazards, including:
- chemical hazards;
- microbiological hazards;
- physical hazards;
- accidents;
- vectors.
The effects of environmental exposures on health depend on the social settings in which the exposures occur and on individual behaviours. Behavioural risk factors are sometimes
Main issues
The rational development of a health policy uses inputs from EBD studies in the following ways:
· EBD studies help to identify priority actions in health and the environment, as well as the effectiveness, cost-effectiveness, and social and ethical implications of an intervention;
· EBD studies allow policy actions or interventions to be based on estimated health gains, rather than on “safe environmental levels” of the risk factor alone;
· EBD studies help to assess the performance of a country;
· EBD studies can identify high-risk populations;
· EBD studies allow research to be prioritized.
closely related to physical risk factors (e.g. hygiene is related to sanitation) and modify the health impacts of the physical risk factors. Indeed, the specific contributions of related behavioural and physical risk factors sometimes cannot easily be separated, as for example in the risk factor “water, sanitation and hygiene”.
As the primary aim of EBD estimates is to inform policy, the assessment of risk factors that are most directly relevant to policy would be the most useful. A representation of how the risk factors affect policy options, such as options for energy policy, transportation policy, or emission-reduction policy, would be ideal. Currently, evidence is generally compiled around media and agents (e.g. air quality, food), and EBD assessments focus on these categories because the data are accessible. Examples of major environmental risk factors, their categories, the types of hazards they carry and their overlaps are represented in Figure 2.1.
Figure 2.1 Environmental hazards and risk factors
Policy scenarios are linked to multiple distal causes and are therefore more complicated to assess (Murray & Lopez, 1999a). For example, transportation scenarios are linked to air pollution, accidents and noise, and the health effects of reduced physical activity. Similarly, energy policies may be linked to a variety of risk factors, including air pollution, accidents, radiation, water pollution etc., according to the selected technology.
Air pollution
Microbiological hazards Chemical hazards
Accidents Physical hazards
Noise
EMF UV
Media
Specific agents
Vectors Road
traffic acci- dents
General environment Occupational environment
Type of hazard
Risk factors:
Lead
Food safety
Pesticides Water supply, sanitation
& hygiene
Arbo- viruses
and proto- zoans
High-risk natural environments, such as wetlands Water resources Agricultural
environments Ion.
Ion.: Ionizing radiation EMF: Electromagnetic fields
2.3 Attributable risk
BoD assessments strive to provide a quantitative answer to the question, “how big is this particular health problem?”. This requires two steps: establishing an appropriate measure of health status, and deciding how much ill-health can be attributed to a particular risk factor. Measures of health status are discussed in Section 4. In this section, we review attributable risk, and how it is applied to EBD calculations. More detailed descriptions of attributable risk can be found in the literature (WHO, 1993; Rothman & Greenland, 1998).
Attributable risk is one of the fundamental concepts underlying BoD assessments, and it involves the ideas of attribution and causal inference. Sometimes we can say, with confidence, that an occurrence of disease is due to a particular environmental exposure.
Mesothelioma, for example, is a neoplasm of the lung that is seldom caused by anything other than asbestos. But this is an exception, and in most cases there are many possible causes of a disease. Gastroenteritis may result from drinking contaminated water, but it may also be caused by toxins in food, or pathogens spread from hand to mouth. Because most diseases have multiple origins, attributable risk cannot be applied at the level of an individual. If a child scores poorly on IQ tests this may be caused by exposure to lead, but in the case of an individual child it may be impossible to exclude other, plausible explanations (such as social disadvantage).
When individuals are grouped together, the task of attribution is more straightforward. If two groups are alike in all important respects, except that one group has been exposed to a factor of interest, then any difference in disease rates between the two groups is said to be attributable to (caused by) the exposure. The disease rates can be equated to risk levels for the populations, and the attributable risk for the factor calculated from the difference (Figure 2.2). Although such data do not distinguish between individuals who fell ill because they were exposed to the factor and those who fell ill from other causes, the data can be used to deduce what fraction of the total BoD would have been avoided if the exposure had not occurred.
Main issues
Environmental risk factors can be categorized according to:
- the media carrying the hazards;
- individual risk factors;
- the nature of the hazard;
- the general environmental or occupational environment.
EBD assessments that are related to policy scenarios would be most useful, but such assessments are complex.
Figure 2.2 In general, attributable risk is the difference in risk levels for “exposed”
and “not exposed” populations
The proportion of the total burden of disease that is due to exposure to an environmental risk factor is called the attributable fraction for that risk factor. This gives an indication of how much ill-health might be avoided if exposure to the risk factor did not occur. What often causes confusion is that the attributable disease fractions for risk factors may not add up to unity. To understand why this might be so, consider an analogy in which the causes of disease are different routes that can be taken to reach a common destination, such as the different ways of travelling to work. An imaginary commuter has four ways of getting to work:
1. Cycle to the station, catch the train, and cycle to work.
2. Drive to the station, catch the train, and walk to work.
3. Walk to the bus stop, and catch the bus to work.
4. Drive to work and walk from the car-park.
It is assumed that the four routes are chosen with equal frequency (i.e. 25%), and that there is no replacement possible (if the trains are on strike one day, then the commuter simply misses a day of work). Also, no means of transportation by itself is sufficient to take the commuter all the way from home to work - they need to be used in combination - and none is necessary (i.e. without it, the commuter would never get to work).
If the methods of transportation are described as “component causes”, the four combinations of transportation that get the commuter to work can be diagrammed as follows:
1 2 3 4
where B = taking the bus, C= cycling, D = driving, T = taking the train, W = walking.
C T T
D W W B D W
Not exposed
Attributable risk
Risk level
Exposed
From the diagrams, it is simple to calculate the proportion of trips to work that use a particular transportation method. Walking, for example, is used in combinations 2, 3 & 4, or 75% (25%+25%+25%) of the trips. Driving is used in combinations 2 & 4, or 50%
(25%+25%) of the trips. The corresponding values for taking the bus, cycling and taking the train are 25%, 25% and 50%, respectively. Each value represents the proportion of trips to work that would be prevented if that particular form of transportation were not available.
The fact that these fractions add up to more than 200% does not mean that more than 200%
of trips to work could be prevented by avoiding all exposures to buses, cycling, train travel etc. Rather, each fraction describes the change that would occur if that particular exposure were altered, while all others stayed the same.
Because these fractions are interdependent, they cannot simply be added up. Once one of the transportation methods has been removed, two things change. First, the frequency of the outcome (getting to work) is altered; and second, the proportions for the remaining transportation methods alter. For example, if bus travel were impossible, there would not only be 25% fewer trips to work, but the proportions for the remaining forms of transportation would also change. Since cycling would be used only in trip combination 1, it would be used in 33% of the trips (not 25%, as before), and trains, driving and walking would each be used in 67% of the trips. Another example of competing and interacting risks is given in Box 2.2.
Box 2.2 Competing risksa
When analysing multiple risks it is natural to think that attributable disease fractions for individual risk factors should all total to 1.0 (or 100%). There are several reasons why this is not so in practice, and the disease fractions for individual risk factors can total to more than, or less than, 100%. Furthermore, the disease fractions may not be strictly additive. Examples of these situations are given below.
Perhaps, it is most easily seen that disease fractions for known risk factors might total less than 100% of the disease burden, given that our knowledge might be incomplete. In other words, if we have not identified all the risks associated with a disease, we may not be able to attribute 100% of the disease burden.
But surely, most people might say, disease percentages due to multiple attributable risks could never add to more than 100%? (After all, how can we prevent more disease than there actually is?). To illustrate how this can be so, consider a hypothetical situation of 1000 annual deaths from auto accidents along a dangerous stretch of highway. Studies show that the deaths would be reduced by 20% if headlight use was required during the day, 40% through stricter speed limits, 50% by installing more stop-lights, 90% by installing speed bumps, 98%
by having a police officer accompany each car, and so on. Clearly, the total, 298%, is open- ended and reflects the detail with which we understand the problem, and our ingenuity in finding ways to deal with it. In this way, the diseases that could be prevented by removing various risk factors can add to more than 100%.
Another factor is that many important risk factors do not create disease cases by themselves, but act in conjunction with other hazards or conditions. Also, certain hazardous features of a risk factor can be compensated by other protecting circumstances. In other words, such risks factors are not usually completely independent, and changes in one will affect the others.
As a result, the attributable disease fractions for multiple risk factors may not be strictly additive. In the highway safety example, if we can save 100 lives by requiring daytime use of headlights, or 200 lives through stricter speed limits, could we save 300 lives by doing both? No, because once one intervention is implemented the overall situation changes and the remaining potential benefits of the other risk factors will be reduced. In this case, many of the 200 people whose lives might be saved in a speed-limit campaign might also have been saved by a campaign to use daytime headlights. Depending on the degree of non- independence, the remaining benefit of requiring daytime use of headlights might only be 50, for example. The total number of lives saved by instigating both campaigns would then be only 250, not 300.
a Adapted from: Smith et al. (1999).
How could we apply this example to burden of disease assessments? The “destination” in this instance is generally a state of impaired health (measured in units such as DALYs), and the various forms of “transportation” are factors that combine in different ways to move an individual from good health to poor health. In other words, the fraction of a disease that is attributable to a risk factor tells us what health losses would be avoided if the risk factor were eliminated. (Commonly, the risk factor cannot be eliminated altogether and so the
“avoidable” risk is calculated, in which case the counterfactual is not “no exposure”, but
“minimum achievable exposure”).
For example, acute respiratory illness in childhood may result from one or more environmental factors, including second-hand smoke, indoor air pollution from other sources, crowding in the home, and attendance at group child-care. There are also factors that affect susceptibility to environmental threats, and some of these are known and can be measured, such as infant feeding (bottle or breast) and past history of chest illness. None of the environmental factors is a sufficient cause of acute respiratory illness (e.g. not all children whose parents smoke become ill), and none is a necessary cause (e.g. acute respiratory illness does not occur exclusively among children of smoking parents). Rather, it is a combination of environmental factors and causes of susceptibility that provides a sufficient cause of illness. Each individual case results from the combined effects of a number of environmental causes (not all identified). Removing one environmental factor (such as second-hand smoke) will reduce the frequency of acute respiratory illness, and may alter the fractions of the disease burden that may be attributed to the remaining risk factors.
2.4 Limitations of EBD studies
The limitations of EBD assessments should be considered in context, and include:
Important aspects of risk are not included
Priorities are not decided on the basis of numbers alone. Some risks pose specific environmental hazards that influence what individuals and populations regard as important.
Factors such as equity, uncertainty, dread and degree of control can all play a role in shaping people’s attitudes. For example, the priority given to controlling spray drift (pesticides drifting into residential areas) is likely to be influenced more by the involuntary and inequitable features of exposure, rather than the resulting burden of disease. It is possible to weight EBD calculations in response to perceptions of different risks (e.g. Box 2.1), but this is a matter to be decided at local level.
EBD does not account for benefits other than health gain
This is important with environmental modifications that may have other kinds of social benefits. An intervention to protect water supplies, for example, can improve food
Sensitivity analyses can test the implications of different choices
BoD calculations involve judgements about standard life expectancy, severity (disability) weights, age weighting, and discounting over time. To see if these judgements make a difference to priorities for action, EBD data should be calculated using different assumptions.
The underlying causal processes may be, necessarily, simplified
Two aspects of complexity are relevant for EBD estimates. First, people are exposed to a complex mix of factors in the environment, and the interactions between these factors are often not well-understood and cannot be modelled satisfactorily in BoD calculations. For example, the effects on the lung of air pollutants such as ozone and NO2 are greater when these factors are present together than when they are inhaled separately (in experimental chambers, for instance). Nevertheless, EBD estimates often treat pollutants individually, because little is known about joint effects. Second, BoD studies do not yet allow for co- morbidity to be estimated, and outcomes tend to be dealt with singly, similar to the approach used with exposures. Yet when diseases occur together, the combined impact on the level of disability is likely to be different from the impact of a single disease.
Parameters that can be measured rapidly may be favoured over those that are difficult to measure
This caveat is relevant to all quantitative measures, but it is important that exposures and outcomes are not overlooked simply because the “right kind” of data are lacking.
Examples range from simple health problems that are poorly documented (e.g. back pain), to the health effects of social and cultural dislocation.
2.5 What are the links between EBD assessments and policy-making?
In section 2.1, several links between EBD assessment and policy were highlighted.
Basically, EBD assessments provide an important input to the development and evaluation of policies in the health sector and to activities of other sectors that directly manage or influence the determinants of health.
The DPSEEA (Driving-force – Pressure – State – Exposure – Effect – Action) framework (Figure 2.3; Kjellström & Corvalán, 1995;Corvalán, Briggs & Zielhuis, 2000) provides a hierarchical model that describes the actions of various causes that act, more or less directly, on health outcomes from environmental or related behavioural conditions. In addition, it displays the various levels of actions that can be taken to reduce health impacts.
a Adapted from WHO (1997).
The DPSEEA framework can be developed for most environmental risk factors and their associated health outcomes. This qualitative model can be developed further into a quantitative one if numerical functions are ascribed to the linkages between the various positions. It is then called a “causal web” and serves as a basis for EBD assessment (Murray & Lopez, 1999a). Ideally, policies act to reduce the causes of disease, and the subsequent impact of policies on the disease burden can be measured from the change in disease burden. In practice, however, it is often difficult to attribute changes in disease burden to particular policy interventions. Reasons include the lags that occur between policy implementation, exposure reduction and disease incidence, and the multicausal nature of disease. For example, lead in the environment is influenced by many factors, and the impact of regulations to remove lead from gasoline and paint on the rates of anaemia and high blood pressure may be difficult to discern (Figure 2.4).
Each of the volumes on risk factors start with the causal web, which constitutes a basis for Economic and social policies
Cleaner technologies Management of
hazards Improvement of
environmental state/safety
Education Increased awareness
Medical treatment Figure 2.3 Conceptual framework for environmental health assessment,
management and evaluation
Actions Driving force
Developmental
Population
growth Economic
development Technology
Pressure Distal cause
Production pattern
State Proximal Exposure Physical/patho- physiologicalcause
Effect Outcome
Well-being Morbidity Mortality Consumption
pattern Behaviour
Natural hazards
Availability of resources
Level of pollution
Personal exposure scenario
Absorbed
dose Pathogen
intake
Figure 2.4 Causal web for lead exposure
Use of leaded gasoline;
traffic density Leaded pipes for water supply in the community
Distal causes
Lead
concentration in air and dust;
personal
Measurable body burden of lead (blood lead level) Lead in food
Proximal
Physiological &
pathophysiological
causes Outcome
Neurological effects IQ loss Increased blood pressure Anaemia Gastrointestinal effect Lead
concentration in drinking water
Occupational exposure (lead in air, dust and occupational environment) Industrial activity
Use of leaded ceramics or cans containing food and drinks
Use of lead in cosmetics and folk remedies Leaded paints in homes
Quantified exposure-risk relationships
Relationship that can be quantified/ modelled on the basis of the literature
Action
Use of unleaded gasoline Removal of leaded water pipes Regulations for ceramics, paints etc.
Action
Non-quantified link
Main issues
· An EBD assessment is an input to rational policy making in environmental health. Policies and interventions can modify health impacts from environmental risk factors at various levels.
· The DPSEEA (Driving-force – Pressure – State – Exposure – Effect – Action) framework is a hierarchical model linking measurable indicators to environmentally-caused diseases.
Such a model can be developed quantitatively, and is then called a “causal web”.
This example of a causal web for lead highlights the relationships between causes that can be quantified and used to estimate the causal web, and how policy actions impact the framework. In this case, the available knowledge on quantitative information allows not only the disease burden caused by environmental exposure to lead to be estimated, but also a prediction of health gains according to policy actions. More detailed information on lead exposure is provided in the volume on lead.
2.6 Data and indicators for EBD assessments
Many countries collect a wealth of health and environmental data, but do not process them into quantitative information about the environmental health impacts on the population.
The statistical data and trends may be used to formulate policies that aim at not exceeding
“safe levels” of environmental pollutants, but this alone is not always directly relevant for policy-making or its evaluation. The key information that is needed to evaluate environmental health policies, or to assess the impact of an intervention, is an estimation of the health impacts.
With new methods for EBD assessment becoming available (such as those presented in this guide), the collection of environmental health data can be designed in such a way that the data can be directly processed into quantified health impacts. Selected indicators can thus be fitted into a framework for data assessment and management that includes collecting and processing the data, and using the information to inform policy and to provide feedback for evaluation. This would increase the value of the environmental data to policy-making in both the environmental and health sectors. A set of such indicators can be converted into disease burden estimates at global level (WHO, 2002), or can be used in guiding an assessment of the EBD at national or local level. These indicators are summarized in Table 2.1.
The selection and collection of data, and the processing of indicators into relevant public health information according to a comprehensive framework, will require close cooperation between the health and environmental sectors, since data on the quality of the environment are generally collected by the environmental sector, whereas environmental health activities are concentrated in the health sector. An EBD assessment does not necessarily require that additional data be collected, but rather a targeted collection of specific data that can be processed into information about the health impacts. Thus, although additional resources are not necessarily required, effective intersectoral collaborations are needed to translate information into health-promoting actions.
Table 2.1 Indicators that can be processed into disease burden estimates using existing methodsa
Area Indicatorb Related diseasesc or
outcomes
Availability of indicator data
Water, sanitation and hygiene.
Water supply coverage.
Sanitation coverage.
Diarrhoea;
total fraction of schistosomiasis, hookworm disease, ascariasis, trichuriasis, trachoma.
Coverage at country level, in Water Supply and Sanitation Assessment 2000.
Ambient air pollution. Annual mean concentrations of particulate matter (PM10
and PM2.5).
Mortality from
cardiopulmonary disease.
Measurements almost only available in developed countries;
global exposure modelled by the World Bank; modelled and measured data cover more than 3000 cities.
Climate change:
extreme temperatures.
Change in daily temperature distributions.
Cardiovascular diseases.
Respiratory diseases (mortality only).
Based on current data and generated by climate models for missing data or projections.
Climate change:
warming.
Change in monthly temperature.
Diarrhoea. Based on current data and generated by climate models for missing data or projections.
Climate change: food production.
Change in temperature, rainfall and CO2.
Malnutrition. Generated on the basis of current data and climate models; food availability is modelled.
Climate change:
coastal floods.
Sea level rise and frequency of coastal floods.
Deaths and injuries. Generated on the basis of current data and climate models; model for frequency of coastal floods.
Climate change:
inland floods, landslides.
Monthly rainfall exceeding the 1-in-10 year limit.
Deaths and injuries. Generated on the basis of current data and climate model.
Climate change:
vector-suitable environments.
Average monthly temperature, minimum annual temperature, average rainfall, resulting in area suitable for malaria transmission.
Malaria.
Dengue.
Climate parameters based on observed and modelled data, and a vector model generates the transmission potential of malaria and dengue.
Indoor air pollution. Household solid fuel use. Strong evidence:
- acute respiratory infections;
- lung cancer;
- chronic obstructive pulmonary disease.
Moderate evidence:
- asthma;
- cataract;
- tuberculosis.
Household solid fuels database and model for predicting fuel use.
Malnutritiond. Stunting.
Wasting.
Underweight.
Child mortality.
Contribution to acute respiratory diseases.
Contribution to malaria.
WHO Global Database on Child Growth and Malnutrition, includes 400 000 measurements that are representative of 85% of the world's children.
Main issues
· Environmental indicators that can be directly processed into health impact information may be of greater value for the environmental and health sectors than indicators that cannot be directly processed.
· Fitting indicators into a comprehensive framework of environmental assessment and management allows the state of the environment to be measured, as well as the health impacts and health gains of interventions.
What information is routinely collected?
To date, only a limited number of EBD methods have been developed for risk factors and the outcomes caused by those risk factors. However, information from currently assessed indicators could be converted into new EBD assessments. In this section we therefore provide an overview of the types of indicators that are currently being collected.
Countries, and national and international agencies, routinely collect information on indicators at different levels of aggregation, both on the environment and on health. The use and interpretation of indicators have been described as crucial links in the data-to- decision-making chain: measurements produce raw data; data are aggregated and summarized to provide statistics; statistics are analysed and re-expressed in the form of indicators; and indicators are then fed into the decision-making process (Wills & Briggs, 1995). An environmental health indicator can be understood as a measure which summarizes in easily understandable and relevant terms some aspect of the relationship between the environment and health that is amenable to action. It is a way of expressing the scientific knowledge about the link between environment and health in a form that can help decision-makers make more informed and more appropriate choices (Corvalán, Briggs
& Zielhuis, 2000). The conversion of indicators into a summary measure of population health using a common metric, such as a DALY, provides additional input to decision- making. The following sections provide a summary of some types and sets of indicators.
Environmental indicators
An environmental indicator has been described as “a measurement, statistic or value that provides a proximate gauge or evidence of the effects of environmental management programs or the state or condition of the environment” (USEPA, 1994). As a result, the majority of environmental indicators developed so far describe the environment without any explicit or direct implications for health. Examples include indicators of atmospheric emissions, surface-water quality, designated areas, or threatened wildlife species.
However, although the data are not collected specifically for health-related purposes, they can be converted into health measures in the framework of EBD assessments. In the context of human health, we are mostly concerned with indicators that measure human exposure to potential health risks, which allows the health impact of environmental risk factors to be evaluated.
Health indicators
The majority of health indicators describe the status of, or trends in, health, without any direct reference to the environment. Examples include simple measures of life expectancy, and cause-specific mortality rates where no attempt has been made to estimate health outcomes that are attributable to the environment. Health indicators have been used extensively to monitor the health of populations, both by international agencies at global, regional and country levels, and by countries at national and subnational levels. But many health outcomes can be related to environmental risk factors (e.g. disease related to the environment). Such a disease-oriented approach provides a means of monitoring and assessing the health outcomes of a wide range of environmental exposures. For example, the WHO Environmental Health Indicators Project has identified sets of indicators that focus on health outcomes and their associated risk factors (WHO, 1999). Summary measures of population health, e.g. expressed in DALYs, combine several health indicators into a single metric that characterizes population health, and describes the impacts of intervention.
Environmental health indicators
Environmental health is concerned with environmental factors that influence or directly affect human health (either positively or negatively). An environmental health indicator can thus be defined as “an expression of the link between environment and health, targeted at an issue of specific policy or management concern and presented in a form which facilitates interpretation for effective decision-making” (Corvalán, Briggs & Zielhuis, 2000). As such, it is more than either an environmental indicator or a health indicator: it is based upon a known or postulated relationship between environmental exposure and health. Examples of environmental health indicators used in water-quality studies are provided in Table 2.2.
Table 2.2 Environmental health indicators within the DPSEEA framework:
an example using microbiological contamination of water
Descriptive indicator Action indicator Driving force Level of poverty in the community. Expenditure on water and sanitation
improvements.
Pressure Percent of households without a safe drinking-water supply.
Number of unserved households provided with clean water supply per year.
State Coliforms in water. Extent of water-quality surveillance and water treatment.
Exposure Percentage of population exposed to hazardous water contaminants.
Extent of public education programmes on water hazards and treatment in the
